Models

GPT-5.6 model optimized for cost-sensitive workloads

Moonshot's flagship for long-horizon coding and knowledge work — patient with large contexts and especially strong on visual, interactive projects.

Frontier model for complex professional work

GPT-5.6 model that balances intelligence and cost

Anthropic's strongest everyday model. Near-Opus reasoning for coding, agents, and professional work when you want quality at Sonnet cost.

Z.ai's flagship. Strong on practical coding and long-context agentic work — a value pick when you want capable output at a lower price than the closed flagships.

Anthropic's most capable model — reserved for the hardest, highest-stakes reasoning and long-horizon agentic work where you want the strongest possible result.

Anthropic's top reasoning model — careful, honest, and quick to catch its own mistakes, well suited to long, hard problems where correctness is everything.

DeepSeek's flagship. Careful on multi-step work and patient with long prompts — a strong value pick at the top tier.

Google's flagship chat model. Quick and capable across long prompts — a strong everyday pick when responsiveness matters.

Google's flagship image model. Strong on photorealism and natural edits, with quick turnaround — a sharp everyday pick.

OpenAI's flagship image model. Best for typography-heavy art, brand work, and edits where re-runs are expensive.

AssemblyAI's accuracy-focused option. Reliable names, dates, and formatting — pick when getting the words right matters more than speed.